Computer systems have evolved significantly during the last century. Starting from relatively slow, electromechanical data manipulation processors employed primarily by large businesses, present-day computer systems include a broad gamut of markedly higher-speed computation devices ranging from massively parallel processing complexes to highly agile, miniaturizable, portable and interconnectable multiple-function computation engines enjoying far broader distribution and a dramatically richer ensemble of applications than in past. Examples of landmark development areas in the broader supportive genre include transistors; microprocessors; networks, such as the ARPANET and Internet; and video technologies, among others.
One consequence of the dramatic expansion of computer systems has been need for increased memory for storage of computer-related or user-accessible information or data. While ongoing development of larger capacity memories continues to provide improvements in the time required to access memory contents, despite impressive and frequent increases in memory size, substantial performance advantages and improved competitive postures also result from techniques that improve how memory capabilities are employed and accessed.
These kinds of advantages tend to promote scalability, or a capacity to increase or decrease system size, number and/or size of applications that can be simultaneously provided, increasing the number of users who can be serviced at any one time, speed of service and the like. In turn, increased scalability often yields substantial competitive advantage potential, at least in part related to significant improvements in user and system capabilities, for example via dramatic and often continuously modifiable/upgradeable capacities.
As a result of increases in available computing power and speed, coupled with numerous other improvements, there have been sharply and constantly increasing needs for data storage capacity. This is exacerbated at the enterprise level, for example each user may have a copy of a dataset that is already represented elsewhere in a networked or distributed computing system, for a variety of reasons, such as, for example, for business continuity and disaster recovery purposes and for obviating bottlenecking in attempting to access pooled data resources. It is estimated that 75% or more of the data storage used in the average enterprise stores redundant dataset copies. Further, inasmuch as the resultant multiple copies may each contain differently-modified data elements, where the modifications are not coordinated into a central or primary data storage device, issues relating to synchronization of memory contents may occur.
In particular, in nonvolatile memory technologies, i.e., those memory types capable of retaining data without requiring continuous electrical input power, access speeds and capacities have barely kept pace or have largely fallen behind advances in other areas of computing technology, such that data storage is increasingly difficult to manage and is becoming increasingly problematic and time-consuming to archive effectively, and especially to effectuate such in conformance with present-day needs, and also, very notably to achieve such in an accurate manner while providing both robust/enduring data integrity coupled with suitable accessibility. Many types of nonvolatile memory technologies have historically employed magnetically-polarizable media, such as magnetic tape systems, hard drives and floppy disc drives. These types of memories typically employ multiple electromagnetic heads for encoding/writing data via modulation of the polarization state of a portion of the magnetic material, or reading data by sensing the polarization state of that portion of the medium in proximity to the heads. In turn, this requires that the medium be physically translated relative to the heads, which is frequently accomplished via rotation of spindles coupled to the media in conjunction with contemporaneous positioning of the head, e.g., radially, or otherwise, vis-á-vis motion of the media. Consequently, such memory/storage technologies may or often incur latency due to delay involved in physical translation of the medium and/or heads in order to access locations corresponding to specific stored data items.
Mass storage via nonvolatile memory technologies has not achieved any large quantum improvements in roughly fifty years, during the evolution process of spindle-based technologies, including tape drives and other electromechanical approaches such as various disc technologies. Continued reliance on spindle-based nonvolatile data storage has also spawned a legacy of increasingly awkward memory accession and management schemes. Further, the read-write capabilities and limitations associated with such approaches lead to practices including overwriting older versions of data, with a result that prior datasets can be, and often are, destroyed. This destruction may be through deliberate and intentional, or totally inadvertent actions. In turn, destruction of prior datasets may have legally and practical implications.